2 * device_cgroup.c - device cgroup subsystem
4 * Copyright 2007 IBM Corp
7 #include <linux/device_cgroup.h>
8 #include <linux/cgroup.h>
9 #include <linux/ctype.h>
10 #include <linux/list.h>
11 #include <linux/uaccess.h>
12 #include <linux/seq_file.h>
13 #include <linux/slab.h>
14 #include <linux/rcupdate.h>
15 #include <linux/mutex.h>
20 #define ACC_MASK (ACC_MKNOD | ACC_READ | ACC_WRITE)
24 #define DEV_ALL 4 /* this represents all devices */
26 static DEFINE_MUTEX(devcgroup_mutex);
29 * whitelist locking rules:
30 * hold devcgroup_mutex for update/read.
31 * hold rcu_read_lock() for read.
34 struct dev_whitelist_item {
38 struct list_head list;
43 struct cgroup_subsys_state css;
44 struct list_head whitelist;
48 static inline struct dev_cgroup *css_to_devcgroup(struct cgroup_subsys_state *s)
50 return container_of(s, struct dev_cgroup, css);
53 static inline struct dev_cgroup *cgroup_to_devcgroup(struct cgroup *cgroup)
55 return css_to_devcgroup(cgroup_subsys_state(cgroup, devices_subsys_id));
58 static inline struct dev_cgroup *task_devcgroup(struct task_struct *task)
60 return css_to_devcgroup(task_subsys_state(task, devices_subsys_id));
63 struct cgroup_subsys devices_subsys;
65 static int devcgroup_can_attach(struct cgroup *new_cgrp,
66 struct cgroup_taskset *set)
68 struct task_struct *task = cgroup_taskset_first(set);
70 if (current != task && !capable(CAP_SYS_ADMIN))
76 * called under devcgroup_mutex
78 static int dev_whitelist_copy(struct list_head *dest, struct list_head *orig)
80 struct dev_whitelist_item *wh, *tmp, *new;
82 list_for_each_entry(wh, orig, list) {
83 new = kmemdup(wh, sizeof(*wh), GFP_KERNEL);
86 list_add_tail(&new->list, dest);
92 list_for_each_entry_safe(wh, tmp, dest, list) {
99 /* Stupid prototype - don't bother combining existing entries */
101 * called under devcgroup_mutex
103 static int dev_whitelist_add(struct dev_cgroup *dev_cgroup,
104 struct dev_whitelist_item *wh)
106 struct dev_whitelist_item *whcopy, *walk;
108 whcopy = kmemdup(wh, sizeof(*wh), GFP_KERNEL);
112 list_for_each_entry(walk, &dev_cgroup->whitelist, list) {
113 if (walk->type != wh->type)
115 if (walk->major != wh->major)
117 if (walk->minor != wh->minor)
120 walk->access |= wh->access;
126 list_add_tail_rcu(&whcopy->list, &dev_cgroup->whitelist);
131 * called under devcgroup_mutex
133 static void dev_whitelist_rm(struct dev_cgroup *dev_cgroup,
134 struct dev_whitelist_item *wh)
136 struct dev_whitelist_item *walk, *tmp;
138 list_for_each_entry_safe(walk, tmp, &dev_cgroup->whitelist, list) {
139 if (walk->type == DEV_ALL)
141 if (walk->type != wh->type)
143 if (walk->major != ~0 && walk->major != wh->major)
145 if (walk->minor != ~0 && walk->minor != wh->minor)
149 walk->access &= ~wh->access;
151 list_del_rcu(&walk->list);
152 kfree_rcu(walk, rcu);
158 * dev_whitelist_clean - frees all entries of the whitelist
159 * @dev_cgroup: dev_cgroup with the whitelist to be cleaned
161 * called under devcgroup_mutex
163 static void dev_whitelist_clean(struct dev_cgroup *dev_cgroup)
165 struct dev_whitelist_item *wh, *tmp;
167 list_for_each_entry_safe(wh, tmp, &dev_cgroup->whitelist, list) {
174 * called from kernel/cgroup.c with cgroup_lock() held.
176 static struct cgroup_subsys_state *devcgroup_create(struct cgroup *cgroup)
178 struct dev_cgroup *dev_cgroup, *parent_dev_cgroup;
179 struct cgroup *parent_cgroup;
182 dev_cgroup = kzalloc(sizeof(*dev_cgroup), GFP_KERNEL);
184 return ERR_PTR(-ENOMEM);
185 INIT_LIST_HEAD(&dev_cgroup->whitelist);
186 parent_cgroup = cgroup->parent;
188 if (parent_cgroup == NULL) {
189 struct dev_whitelist_item *wh;
190 wh = kmalloc(sizeof(*wh), GFP_KERNEL);
193 return ERR_PTR(-ENOMEM);
195 wh->minor = wh->major = ~0;
197 wh->access = ACC_MASK;
198 dev_cgroup->deny_all = false;
199 list_add(&wh->list, &dev_cgroup->whitelist);
201 parent_dev_cgroup = cgroup_to_devcgroup(parent_cgroup);
202 mutex_lock(&devcgroup_mutex);
203 ret = dev_whitelist_copy(&dev_cgroup->whitelist,
204 &parent_dev_cgroup->whitelist);
205 dev_cgroup->deny_all = parent_dev_cgroup->deny_all;
206 mutex_unlock(&devcgroup_mutex);
213 return &dev_cgroup->css;
216 static void devcgroup_destroy(struct cgroup *cgroup)
218 struct dev_cgroup *dev_cgroup;
220 dev_cgroup = cgroup_to_devcgroup(cgroup);
221 dev_whitelist_clean(dev_cgroup);
225 #define DEVCG_ALLOW 1
232 static void set_access(char *acc, short access)
235 memset(acc, 0, ACCLEN);
236 if (access & ACC_READ)
238 if (access & ACC_WRITE)
240 if (access & ACC_MKNOD)
244 static char type_to_char(short type)
248 if (type == DEV_CHAR)
250 if (type == DEV_BLOCK)
255 static void set_majmin(char *str, unsigned m)
260 sprintf(str, "%u", m);
263 static int devcgroup_seq_read(struct cgroup *cgroup, struct cftype *cft,
266 struct dev_cgroup *devcgroup = cgroup_to_devcgroup(cgroup);
267 struct dev_whitelist_item *wh;
268 char maj[MAJMINLEN], min[MAJMINLEN], acc[ACCLEN];
271 list_for_each_entry_rcu(wh, &devcgroup->whitelist, list) {
272 set_access(acc, wh->access);
273 set_majmin(maj, wh->major);
274 set_majmin(min, wh->minor);
275 seq_printf(m, "%c %s:%s %s\n", type_to_char(wh->type),
284 * may_access_whitelist:
285 * does the access granted to dev_cgroup c contain the access
286 * requested in whitelist item refwh.
287 * return 1 if yes, 0 if no.
288 * call with devcgroup_mutex held
290 static int may_access_whitelist(struct dev_cgroup *c,
291 struct dev_whitelist_item *refwh)
293 struct dev_whitelist_item *whitem;
295 list_for_each_entry(whitem, &c->whitelist, list) {
296 if (whitem->type & DEV_ALL)
298 if ((refwh->type & DEV_BLOCK) && !(whitem->type & DEV_BLOCK))
300 if ((refwh->type & DEV_CHAR) && !(whitem->type & DEV_CHAR))
302 if (whitem->major != ~0 && whitem->major != refwh->major)
304 if (whitem->minor != ~0 && whitem->minor != refwh->minor)
306 if (refwh->access & (~whitem->access))
315 * when adding a new allow rule to a device whitelist, the rule
316 * must be allowed in the parent device
318 static int parent_has_perm(struct dev_cgroup *childcg,
319 struct dev_whitelist_item *wh)
321 struct cgroup *pcg = childcg->css.cgroup->parent;
322 struct dev_cgroup *parent;
326 parent = cgroup_to_devcgroup(pcg);
327 return may_access_whitelist(parent, wh);
331 * Modify the whitelist using allow/deny rules.
332 * CAP_SYS_ADMIN is needed for this. It's at least separate from CAP_MKNOD
333 * so we can give a container CAP_MKNOD to let it create devices but not
334 * modify the whitelist.
335 * It seems likely we'll want to add a CAP_CONTAINER capability to allow
336 * us to also grant CAP_SYS_ADMIN to containers without giving away the
337 * device whitelist controls, but for now we'll stick with CAP_SYS_ADMIN
339 * Taking rules away is always allowed (given CAP_SYS_ADMIN). Granting
340 * new access is only allowed if you're in the top-level cgroup, or your
341 * parent cgroup has the access you're asking for.
343 static int devcgroup_update_access(struct dev_cgroup *devcgroup,
344 int filetype, const char *buffer)
349 struct dev_whitelist_item wh;
351 if (!capable(CAP_SYS_ADMIN))
354 memset(&wh, 0, sizeof(wh));
360 wh.access = ACC_MASK;
380 } else if (isdigit(*b)) {
381 wh.major = simple_strtoul(b, &endp, 10);
394 } else if (isdigit(*b)) {
395 wh.minor = simple_strtoul(b, &endp, 10);
402 for (b++, count = 0; count < 3; count++, b++) {
405 wh.access |= ACC_READ;
408 wh.access |= ACC_WRITE;
411 wh.access |= ACC_MKNOD;
425 if (!parent_has_perm(devcgroup, &wh))
427 devcgroup->deny_all = false;
428 return dev_whitelist_add(devcgroup, &wh);
430 dev_whitelist_rm(devcgroup, &wh);
431 devcgroup->deny_all = true;
439 static int devcgroup_access_write(struct cgroup *cgrp, struct cftype *cft,
444 mutex_lock(&devcgroup_mutex);
445 retval = devcgroup_update_access(cgroup_to_devcgroup(cgrp),
446 cft->private, buffer);
447 mutex_unlock(&devcgroup_mutex);
451 static struct cftype dev_cgroup_files[] = {
454 .write_string = devcgroup_access_write,
455 .private = DEVCG_ALLOW,
459 .write_string = devcgroup_access_write,
460 .private = DEVCG_DENY,
464 .read_seq_string = devcgroup_seq_read,
465 .private = DEVCG_LIST,
470 struct cgroup_subsys devices_subsys = {
472 .can_attach = devcgroup_can_attach,
473 .create = devcgroup_create,
474 .destroy = devcgroup_destroy,
475 .subsys_id = devices_subsys_id,
476 .base_cftypes = dev_cgroup_files,
479 int __devcgroup_inode_permission(struct inode *inode, int mask)
481 struct dev_cgroup *dev_cgroup;
482 struct dev_whitelist_item *wh;
486 dev_cgroup = task_devcgroup(current);
488 list_for_each_entry_rcu(wh, &dev_cgroup->whitelist, list) {
489 if (wh->type & DEV_ALL)
491 if ((wh->type & DEV_BLOCK) && !S_ISBLK(inode->i_mode))
493 if ((wh->type & DEV_CHAR) && !S_ISCHR(inode->i_mode))
495 if (wh->major != ~0 && wh->major != imajor(inode))
497 if (wh->minor != ~0 && wh->minor != iminor(inode))
500 if ((mask & MAY_WRITE) && !(wh->access & ACC_WRITE))
502 if ((mask & MAY_READ) && !(wh->access & ACC_READ))
514 int devcgroup_inode_mknod(int mode, dev_t dev)
516 struct dev_cgroup *dev_cgroup;
517 struct dev_whitelist_item *wh;
519 if (!S_ISBLK(mode) && !S_ISCHR(mode))
524 dev_cgroup = task_devcgroup(current);
526 list_for_each_entry_rcu(wh, &dev_cgroup->whitelist, list) {
527 if (wh->type & DEV_ALL)
529 if ((wh->type & DEV_BLOCK) && !S_ISBLK(mode))
531 if ((wh->type & DEV_CHAR) && !S_ISCHR(mode))
533 if (wh->major != ~0 && wh->major != MAJOR(dev))
535 if (wh->minor != ~0 && wh->minor != MINOR(dev))
538 if (!(wh->access & ACC_MKNOD))